Will the Weather Gods Smile or Frown? Evaluating Monsoon Forecasts* - RBI

Page created by Hugh Murray
 
CONTINUE READING
Will the Weather Gods Smile or Frown? Evaluating Monsoon Forecasts                                                               article

Will the Weather Gods Smile or                                                        At present, the official forecast of the SWM is
                                                                                 given by the IMD, first in April - the First Stage Long-
Frown? Evaluating Monsoon                                                        Range Forecast (FSLRF) - and again in May/June -
Forecasts*                                                                       Second Stage Long-Range Forecast (SSLRF). Skymet, a
                                                                                 private forecaster, releases its preliminary forecast in
                                                                                 April and a revision in May. Apart from these Indian
      The India Meteorological Department (IMD)
                                                                                 agencies, the forecasts of international meteorological
has predicted normal and well distributed South West
                                                                                 organisations, viz., USA’s National Oceanic and
Monsoon rainfall for 2019, which is at odds with the
                                                                                 Atmospheric Administration (NOAA) and Australia’s
forecasts of private and international agencies. This
article drills analytically into past patterns of prediction-                    Bureau of Meteorology (BOM) on El Nino/La Nina
outcome performances of various agencies through                                 and dipole conditions in the Pacific Ocean and Indian
multiple statistical measures to evaluate their forecast                         Ocean, respectively, are also used to predict the
accuracy. The comparative assessment suggests that for                           monsoon conditions in India.
generating macroeconomic forecasts, the use of IMD’s                                  Weather forecasters contend with enormous
second stage long range forecast and the predictions of                          complexities in no small measure, which has
international agencies in conjunction may be appropriate                         catalysed extraordinary progress in the field with
as the preliminary forecasts of IMD and Skymet released                          supercomputers and satellite images enabling high
in April appear to be noisy.                                                     frequency precise measurement of temperature, air
                                                                                 pressure, humidity, wind speed and direction (Silver,
Introduction
                                                                                 2012). India received a significant boost to its weather
      India receives 75 per cent of its annual rainfall
                                                                                 forecasting capacity in 2018 with the Pratyush and
during the South West Monsoon (SWM) season, which
                                                                                 Mihir supercomputers, elevating it next only to Japan,
is spread over June to September. With 65 per cent of
                                                                                 the UK and the US in terms of dedicated high capacity
Gross Cropped Area (GCA) in the country not under any
                                                                                 computing. The IMD’s precise prediction of the
irrigation cover and even different sources of irrigation
                                                                                 cyclone Fani in Odisha in 2019 and resultant disaster
depending on rainfall during the SWM season to build
                                                                                 risk reduction received global acclaim, including from
up storage levels (GOI, 2018), the SWM remains the
                                                                                 the United Nations.
life force of India’s agriculture and for the broader
economy1 (Gulati, et al., 2013). Furthermore, the SWM                                 Evaluating forecast performance in respect of
influences both Kharif (June-September) production                               the SWM is the key motivation of the article. At the
through its temporal and spatial distribution, and Rabi                          outset itself, it is sensitive to the reality that the
(October-March) production through its impact on soil                            science of weather has enabled enormous progress
moisture conditions and storage levels in reservoirs.                            even on long-range forecasts, but no forecast is free
Consequently, SWM forecasts remain critical to every                             of errors. Recognising this caveat, the purpose of
assessment of India’s macroeconomic outlook.                                     the article is to drill analytically into past patterns
*
  This article is prepared by Priyanka Bajaj, D. Suganthi, Rishabh Kumar and     of prediction-outcome performance in order to draw
Atri Mukherjee of the Division of Rural Economics, Department of Economic        some meaningful insights from the IMD’s forecasts.
and Policy Research, Reserve Bank of India. The views expressed in this
article are those of the authors and do not represent the views of the Reserve   Both FSLRF and SSLRF released in April and May,
Bank of India.
1                                                                                have predicted rainfall during June-September 2019 to
 Share of Agriculture and allied sector in india's total Gross Value Added is
15.3 percent (average of last five years).                                       be normal at 96 per cent of the Long Period Average

RBI Bulletin June 2019                                                                                                                15
article                                                                    Will the Weather Gods Smile or Frown? Evaluating Monsoon Forecasts

(LPA) with a model error of ±5 and ±4 per cent,                                 2015)5 and deficient rainfall year (2004), forecast errors
respectively.2 The IMD has also predicted the rainfall                          (actual minus predicted) for FSLRF were negative and
to be well distributed across the states. This, however,                        large in value (Chart 1). SSLRF is more broad-based
is at odds with forecasts of Skymet3, NOAA and BOM4,                            than FSLRF in that it provides rainfall forecasts for
which predict a weak monsoon this year. The rest of                             the four broad geographical regions of India as well as
the paper is structured as follows. Section II presents                         for the country as a whole. The assumed model error
some stylised facts relating to monsoon forecasts by                            for the four regional forecasts is ±8 per cent of the
different agencies. Data, methodology and empirical                             LPA. The SSLRF has out-performed the FSLRF: it could
                                                                                successfully predict the drought of 2015, which was
results are discussed in Section III. Concluding remarks
                                                                                missed out by the FSLRF.
and some policy perspectives are set out in Section IV.
                                                                                     A comparison of IMD’s FSLRF with the predictions
II. Some Stylised Facts
                                                                                of Skymet6 (both are released in April) reveals that
         The IMD introduced SSLRF in 2003, after the                            both the agencies have missed the target in 5 out of
failure of FSLRF to predict the massive drought of 2002.                        the last 6 years. The year 2017 was an exception - the
Out of the 24 years since 1995, FSLRF over-predicted                            forecast was exactly the same as actual for Skymet, and
rainfall in 13 years and under-predicted it in 11 years.                        very close to actual in the case of the IMD’s FSLRF
In the case of the drought years (2002, 2009, 2014 and                          (Chart 2).

2
    The LPA (1951-2000) for the country is 89 cm.
3
 Skymet has predicted a below normal monsoon of 93 per cent of LPA with an error margin of ±5 per cent. Spatially, it expects less than normal rainfall
for all the four regions of the country.
4
 Climate Diagnostics Bulletin (various issues), National Weather Service Climate Prediction Centre, National Oceanic and Atmospheric Administration,
United States Department of Commerce. (Retrieved from https://www.cpc.ncep.noaa.gov/products/CDB/CDB_Archive_html/CDB_archive.shtml )
5
 All-India drought years are announced when the rainfall deficiency is more than 10 per cent of LPA and 20-40 per cent area of the country is under drought
conditions.
6
 Performance of the revised outlook of Skymet could not be evaluated due to paucity of data as it was introduced only in 2017. Incidentally, it did not
revise its prediction for 2017 and 2018.

16                                                                                                                               RBI Bulletin June 2019
Will the Weather Gods Smile or Frown? Evaluating Monsoon Forecasts                                                             article

Regional Forecast Performance                                                   Skymet, which started its regional forecast since 2018,
                                                                                deviation of actual rainfall from forecast is lower for
         Often, the all-India forecast masks the details -
                                                                                North West and Southern regions than for Central and
spatial distribution of rainfall - which assumes crucial
                                                                                North Eastern regions.
importance for the outlook on agricultural production.
For instance, in the year 2007, the overall rainfall in                         Forecast Performance of International Agencies
India was 105 per cent of LPA while the North West                                   The SWM is significantly influenced by El Nino or
region received only 85 per cent of LPA. The largest                            La Nina conditions. During El Nino, sea level pressure
deviations between the IMD’s SSLRF and actual                                   tends to be lower in the Eastern Pacific and higher
rainfall are in relation to the North-Eastern states                            in the Western Pacific. This implies warmer ocean
of India than for the other regions (Chart 3)7. For                             waters in the East and colder ocean waters in the

7
    The region-wise forecast under SSLRF is available only from 2003 onwards.

RBI Bulletin June 2019                                                                                                              17
article                                                                  Will the Weather Gods Smile or Frown? Evaluating Monsoon Forecasts

West. The opposite condition prevails during La Nina.                         1997 despite it being a strong El Nino year, while in
The warming of ocean waters triggers a see-saw                                2014, a negative IOD and El Nino had worked together
in atmospheric pressure between the Eastern and                               to produce deficient rainfall (Karumari, et al., 2001).
Western Tropical pacific, known as Southern Oscillation                       In 2015, however, a positive IOD could not counteract
(SO). Since El Nino and SO are related, the two                               a very strong El Nino and the SWM eventually turned
terms are often combined into a single term ‘El Nino                          out to be deficient.
Southern Oscillation’ or ‘ENSO’. A warm ENSO phase
                                                                                   The predictions of ENSO and IOD are released by
signals El Nino and a cold ENSO phase is related with
                                                                              NOAA and BOM on a monthly and fortnightly basis,
La Nina. El Nino is generally associated with deficient                       respectively.8 It is observed that the NOAA’s forecasts
rainfall in India, whereas the development of La Nina                         of El Nino/La Nina through the spring (March-April)
tends to bring above normal monsoon (Rajeevan and                             tend to be less reliable, while the outlook released in
Pai, 2006).                                                                   June seems to perform better. For instance, in 2017,
     In addition to these developments in the Pacific                         the NOAA’s April outlook predicted a 50 per cent
Ocean, the SWM also depends on the developments                               probability for El Nino; the prediction status was
which take place nearby in the Indian Ocean, known as                         changed to neutral in the June outlook and eventually
Indian Ocean Dipole (IOD) conditions. A positive IOD                          it turned out as a year of normal rainfall. Similarly,
occurs when the western basin of the Indian Ocean                             in the severe drought year 2009, the April outlook
warms up relative to the tropical eastern basin. A                            predicted neutral conditions, which was changed
negative IOD occurs when the reverse takes place. The                         to El Nino in the June outlook by both NOAA and
occurrence of a positive IOD can reduce the impact of                         BOM (with more than 50 per cent probability). In
El Nino, which takes place in the far away Pacific Ocean                      the drought years 2002, 2014 and 2015, both NOAA
and can bring in better rains for India. It is believed                       and BOM predicted El Nino in April and also in June,
that positive IOD had facilitated normal rainfall in                          which turned out to be correct (Chart 4a and 4b).

8
  ENSO Wrap-Up: Current State of the Pacific and Indian Ocean (various issues), Australia Government Bureau of Meteorology. (Retrieved from http://www.
bom.gov.au/climate/enso/wrap-up/archive.shtml)

18                                                                                                                            RBI Bulletin June 2019
Will the Weather Gods Smile or Frown? Evaluating Monsoon Forecasts                                                                    article

     For the 2019 SWM, both NOAA and BOM had                         year, near normal rainfall was received in 1997. On
predicted the chances of formation of El Nino to be                  the other hand, only a moderate El Nino was observed
high in their April outlook, given the above average                 in 2002, which resulted in one of the worst droughts
warm temperature conditions in the Eastern Pacific                   (Table 1).
Ocean. While the BOM had predicted 70 per cent
                                                                     III. Evaluating SWM Forecasts
chance of an El Nino developing this year, the NOAA
had predicted occurrence of El Nino with 60 per cent                       The forecast performance of various models of
probability during the June-August period. In the May                IMD has been examined in the literature (Rajeevan, et
outlook, the BOM has reduced the chances of El Nino                  al., 2004; Kar, et al., 2012; Pandey, et al., 2016; Prasad,
developing in 2019 to 50 per cent, whereas the NOAA                  et al., 2010; Stern, 2008), including the accuracy of the
has revised it upwards to 70 per cent.                               forecast vis-à-vis actual rainfall either for the drought
                                                                     years or for a particular spatial area (Joseph, et al.,
     It has been observed that most of the severe
                                                                     2017; Sagar, et al., 2017). However, comparisons of
droughts in India were influenced by El Nino. In
                                                                     forecast accuracy of IMD vis-à-vis other private and
the period between 1995 and 2018, there have been
                                                                     international agencies has remained a gap.
five extreme deficient rainfall years, four all-India
drought years (2002, 2009, 2014, and 2015) and one                         Accordingly, the forecast accuracy of different
deficient rainfall year (2004). All these five years were            agencies is evaluated through (i) Pearson’s Correlation
associated with El Nino, though the intensity of El Nino             Coefficient (PCC); (ii) Root Mean Square Error (RMSE);
differed across the years. However, development of El                and (iii) success score of predicting extreme events9.
Nino is not a sufficient condition for drought/deficient             9
                                                                      The success score is defined as the proportion of correct forecasts of
                                                                     extreme events to the total actual occurrence of extreme events expressed in
rainfall. Less than half of El Nino events are associated            percentage terms. This includes drought years (2002, 2009, 2014 and 2015),
with deficient rainfall over India (Rajeevan and Pai,                deficient rainfall year (2004) and above normal rainfall years (1998, 2007,
                                                                     2013). The success score value ranges between 0 and 100 per cent where
2006). For instance, despite being a strong El Nino                  higher values denote better forecasting.

RBI Bulletin June 2019                                                                                                                       19
article                                                            Will the Weather Gods Smile or Frown? Evaluating Monsoon Forecasts

                                       Table.1 Vulnerability of Monsoon during El Nino Years

                                             Indian Ocean Dipole           Actual performance                      Monsoon
     Year              Occurrence
                                                    (IOD)                  compared to normal           (per cent departure from LPA)

     1995           Moderate La Nina               Neutral                     Near Normal                          100

     1996                Neutral                  Negative                     Near Normal                          102

     1997          Very Strong El Nino            Positive                     Near Normal                          102

     1998             Strong La Nina              Negative                    Above Normal                          106

     1999             Strong La Nina               Neutral                     Near Normal                           96

     2000                La Lina                   Neutral                    Below Normal                           92

     2001                Neutral                   Neutral                    Below Normal                           92

     2002           Moderate El Nino               Neutral                      Deficient                            81

     2003                Neutral                   Neutral                     Near Normal                          102

     2004             Weak El Nino                 Neutral                      Deficient                            87

     2005                Neutral                   Neutral                     Near Normal                           99

     2006             Weak El Nino                Positive                     Near Normal                           99

     2007             Strong La Nina               Neutral                    Above Normal                          105

     2008                La Nina                   Neutral                     Near Normal                           98

     2009           Moderate El Nino               Neutral                   Severe Drought                          78

     2010             Strong La Nina              Negative                     Near Normal                          102

     2011           Moderate La Nina               Neutral                     Near Normal                          102

     2012                Neutral                  Positive                    Below Normal                           93

     2013                Neutral                   Neutral                    Above Normal                          106

     2014             Weak El Nino                Negative                      Deficient                            88

     2015          Very Strong El Nino            Positive                      Deficient                            86

     2016                La Nina                  Negative                     Near Normal                           97

     2017                Neutral                   Neutral                    Below Normal                           95

     2018                Neutral                   Neutral                    Below Normal                           91
Source: NOAA, BOM and IMD

                                                                      accuracy of all the four agencies (IMD, Skymet,
                                                                      NOAA and BOM) as the predictions are reported
     RMSE =
                                                                      in terms of probability of occurrence by BOM and
     where,     and      are the forecast and actual values,
                                                                      NOAA. For PCC and RMSE, the period of study is from
respectively, and n is the number of observations.
                                                                      1995 to 2018, which is characterised by three distinct
     PCC and RMSE have been used to examine the                       phases – Period 1 (1995 to 2002) with only FSLRF;
forecast accuracy of IMD and Skymet as the predictions                Period 2 (2003 to 2012) with both FSLRF and SSLRF
and actual values are reported as percentage of LPA.                  and Period 3 (2013 to 2018) with FSLRF, SSLRF and
Success score has been used to assess the forecast                    Skymet forecasts.

20                                                                                                              RBI Bulletin June 2019
Will the Weather Gods Smile or Frown? Evaluating Monsoon Forecasts                                                                               article

Analysis of Results                                                                         Table 3: Comparison of RMSE of Rainfall
                                                                                                      Forecasts and Actual
     The PCC results reveal that the correlation
                                                                                    Root Mean Square Error       FSLRF and   SSLRF and         Skymet and
coefficient between the IMD’s FSLRF and actual                                                                     actual      actual             actual

rainfall for Period 1 is negative (-0.36) and insignificant                         Period 1 (1995 to 2002)        10.14             -             -

(Table 2). For Period 2, the correlation coefficient of the                         Period 2 (2003 to 2012)        8.57         8.16               -

FSLRF and SSLRF with actual rainfall is negative (-0.21)                            Period 3 (2013 to 2018)        7.03         5.99              8.71
                                                                                   Source: Authors’ estimates
and positive (0.12), respectively, and both turn out to
be insignificant. In Period 3, the correlation coefficient                         predict all-India droughts (2002, 2009, 2014 and 2015),
between the FSLRF and actual rainfall improves to                                  deficient rainfall (2004) and above normal rainfall
positive and is insignificant (0.45). For the same                                 (1998, 2007 and 2013). Incidentally, it got predictions
period, the correlation coefficient of SSLRF improves                              correct for near normal monsoon only 38 per cent
substantially to 0.64 though it remains insignificant.                             of the time. Similarly, Skymet also failed to forecast
The correlation coefficient of Skymet prognosis and                                extreme events. Nevertheless, the SSLRF nailed down
actual rainfall is positive and insignificant (0.41).                              the 2015 drought and its probability of predicting near
Clearly, none of the forecasts are significantly correlated                        normal monsoon is over 44 per cent. Contrastingly,
with the actual rainfall data across the entire period of                          international agencies NOAA and BOM have been
study to arrive at a conclusive inference.10                                       relatively more successful in forecasting extreme
                                                                                   rainfall years, which generally coincide with El Nino
      The RMSE results are similar to the correlation
                                                                                   and La Nina conditions. This is corroborated by higher
results. The RMSE of the IMD’s FSLRF is higher
                                                                                   values of success score for international agencies
than the SSLRF across time periods. in Period 3,
                                                                                   compared to IMD and Skymet (Table 4). Among
the RMSE of both FSLRF and SSLRF improves though
                                                                                   national forecasts, the success score is higher for the
it still remains above the model error of 5 and 4 per
                                                                                   IMD’s SSLRF than FSLRF and Skymet.
cent, respectively. The RMSE of Skymet is above both
the IMD’s FSLRF and SSLRF (Table 3).                                                     The box plot of the spread of forecast errors
                                                                                   (actual minus forecast) of IMD (FSLRF and SSLRF)
     These results are reflected in the prognosis
                                                                                   and Skymet shows interesting variations (Chart 5).
of extreme events like droughts and above normal
rainfall. For instance, the IMD’s FSLRF failed to                                  The spread is higher for the IMD’s FSLRF in Period 1
                                                                                   which has gradually reduced in subsequent periods.
      Table 2: Comparison of PCC between Rainfall                                  In Period 2, as expected, the error spread of the
                  Forecasts and Actual                                             SSLRF is lower than that of the FSLRF. In Period 3,
                                FSLRF and       SSLRF and       Skymet and
 Time period/ Indicators
                               actual (pval)   actual (pval)    actual (pval)
                                                                                   although the spread is higher for SSLRF as compared
 Period 1 (1995 to 2002)       -0.36 (0.37)           -                -           Table 4: Success Score of Various Forecasting Agencies
 Period 2 (2003 to 2012)       -0.21 (0.57)     0.12 (0.74)            -                       During Extreme Rainfall Years
 Period 3 (2013 to 2018)       0.45 (0.37)      0.64 (0.17)      0.41 (0.42)        Agencies/Month of forecast     March     April       May        June
Source: Authors’ estimates                                                                     NOAA                 57.14    57.14       71.43     57.14
10
  It is recognised that the period wise sample size is low, due to the paucity                 BOM                  71.43    71.43       71.43     71.43
of data for various agencies. Nevertheless, even for the entire period from                     IMD                   -      0.00          -       16.67
1995 to 2018, the PCC of FSRLF (-0.07) is negative and insignificant. Similarly,
the PCC of SSLRF (2003-2018) is positive and insignificant (0.34). These                      Skymet                  -      0.00          -             -
results are further corroborated by RMSE results and Charts 1 & 2 of the           Note: ‘-’ represents not applicable.
Stylised facts.                                                                    Source: NOAA, BOM, IMD and Skymet and author's estimate.

RBI Bulletin June 2019                                                                                                                                       21
article                                                  Will the Weather Gods Smile or Frown? Evaluating Monsoon Forecasts

                                                                  As mentioned earlier, the IMD also provides
                                                            forecast of regional distribution of rainfall (North
                                                            West, Central, East and North East and Southern
                                                            Peninsula) in the SSLRF since 2003. The error plots
                                                            (difference between actual and prediction) show that
                                                            there is no systematic pattern of prediction errors
                                                            across the regions (Chart 6). However, the RMSE is
                                                            highest for the Southern Peninsula (13.24), followed
                                                            by East and North East (12.54), Central India (11.18)
                                                            and North West (9.42). The forecasts tend to over-
                                                            project rainfall in the East and North East and Southern
                                                            peninsula in 77 per cent and 50 per cent of the time,
                                                            respectively. The forecasts have been correct for North
                                                            West nearly 50 per cent of the time and, hence, the
                                                            low RMSE.

                                                            IV. Conclusion
to the FSLRF, the latter has an outlier which makes              There is no significant correlation between the
it less credible. Likewise, the error spread of Skymet      projected rainfall (IMD and Skymet) and actual rainfall
forecast is higher than the IMD’s SSLRF. Clearly,           in India. While none of the forecasts are close to the
in Period 3 the error spread for SSLRF has reduced          actual, the performance of the IMD’s SSLRF is better
from Period 2; however, since the mean is above the         than FSLRF and Skymet. Both IMD and Skymet have
median in the former, SSLRF seems to over-predict the       failed to predict drought and excess rainfall in most
rainfall.                                                   of the cases. Nevertheless, the SSLRF nailed down the

22                                                                                                   RBI Bulletin June 2019
Will the Weather Gods Smile or Frown? Evaluating Monsoon Forecasts                                                   article

2015 drought and its probability of predicting near-                 the Indian Monsoon Rainfall and ENSO’, Geophysical
normal monsoon has been reasonable and higher                        Research Letters, 28.
than FSLRF and Skymet. In contrast, the predictive                   Monsoon Report (various issues), Indian Meteorological
power of the international agencies, viz., BOM and                   Department (IMD), Government of India.
NOAA in forecasting extreme rainfall (which generally                Pandey, D.K., Rai, S., Sahai, A.K., Abhilash, S. and
coincides with the El Nino and La Nina conditions) is                Shahi, N.K. (2016), ‘Prediction and Error Growth in the
much better than that of the IMD. The comparative                    Daily Forecast of Precipitation from the NCEP CFSv2
assessment of all forecasts suggests that for generating             over the Subdivisions of Indian Subcontinent’, Journal
macroeconomic forecasts, the use of IMD’s SSLRF and                  of Earth System Science, Vol 125(1), 29-45.
the predictions of international agencies like NOAA                  Prasad, K., Dash, S.K., and Mohanty, U.C. (2010), ‘A
and BOM in conjunction may be appropriate as the                     Logistic Regression Approach for Monthly Rainfall
preliminary forecasts of IMD (FSLRF) and Skymet                      Forecasts in Meteorological Subdivisions of India based
                                                                     on DEMETER Retrospective Forecasts’, International
released in April appear to be noisy.
                                                                     Journal of Climatology, Vol 30, 1577-1588.
References
                                                                     Rajeevan, M. and Pai, D.S. (2006), ‘On the El Nino-
Government of India (2018), Agriculture at a Glance,                 Indian Monsoon predictive relationships’, Geophysical
Directorate of Economics and Statistics, Ministry of                 Research Letters, Vol. 34.
Agriculture and Farmers Welfare, New Delhi.
                                                                     Rajeevan, M., Pai, D.S., Dikshit, S.K., and Kelkar, R.R.
Gulati, A., Saini, S., and Jain, S. (2013), ‘Monsoon 2013:           (2004), ‘IMD’s New Operational Models for Long-
Estimating the Impact on Agriculture’, Working Paper                 Range Forecast of Southwest Monsoon Rainfall over
No.269, Indian Council For Research on International                 India and their Verification for 2003’, Current Science,
Economic Relations.                                                  Vol 86(3), 422-431.
Joseph, M., Murugan, E. and Hemalatha (2017),                        Sagar, K.S., Mrudula, G., Kumari, K.V., and Rao, S.V.B
‘Forecast Verification Analysis of Rainfall for Southern             (2017), ‘Verification of VARSHA Rainfall Forecasts
Districts of Tamil Nadu, India’, International Journal               for Summer Monsoon Seasons of 2009 and 2010’,
of Current Microbiology and Applied Sciences, 6: 299-                International Journal of Current Research Review, Vol
306.                                                                 9(21), 24-36.
Kar, S.C., Acharya, N., Mohanty, U.C., and Kulkarni,                 Silver, N. (2012), ‘The Weatherman is not a Moron’
M.A. (2012), ‘Skill of Monthly Rainfall Forecasts                    New York Times dated September 7. (Retrieved from
over India using Multi-Model Ensemble Schemes’,                      https://www.nytimes.com/2012/09/09/magazine/the-
International Journal of Climatology, Vol 32, 1271-                  weatherman-is-not-a-moron.html )
1286.                                                                Stern, H. (2008), ‘The Accuracy of Weather Forecasts for
Karumuri, A., Guan, Z. and Toshio, Y. (2001), ‘Impact of             Melbourne, Australia’, Meteorological Applications,
the Indian Ocean Dipole on the relationship between                  15: 65-71.

RBI Bulletin June 2019                                                                                                    23
You can also read